The usefulness of autologous bone marrow transplantation (ABMT) in cancer therapy is limited by uncertainty over the contribution of residual malignant cells in harvested "remission" marrow to recurrence of the disease. Human marrow progenitor cells will be marked with a retrovirally introduced gene (NeoR) to assess the potential of reinfused marrow to contribute to relapse in patients with acute myelogenous leukemia (AML). If the marked cells do contribute to disease recurrence, a second hypothesis will be tested, namely, that marrow purging can remove the malignant blasts before infusion. Information generated by this study will provide insight into the mechanism(s) of relapse after ABMT and an estimate of the value of pretransplant marrow purging. Since retrovirus-mediated gene transfer marks normal as well as leukemic marrow progenitor cells, the information obtained will also be relevant to the biology of marrow ontogeny in man, making it possible, for example to test the contribution of autologous marrow rescue to long-term hematopoietic reconstitution. The final aim is to investigate whether ex vivo treatment of marrow cells with available growth factors will accelerate marrow recovery post-transplantation. If it does, then this approach would simplify the practice and increase the safety of ABMT, particularly if purging - which retards subsequent marrow recovery - proves to be both necessary and effective. If growth factor pretreatment also increases the efficiency and durability of gene transfer, then it will be possible to design protocols for the transfer of therapeutic genes into long-lived marrow progenitor cells. These therapeutic protocols would have wide-ranging application in the treatment of malignant and nonmalignant disease.